BaxBench - No Security Reminder - Correct: leaderboard
BaxBench has models generate complete backend applications from a specification across many web frameworks; functional tests check correctness and security exploits check for vulnerabilities. This board is pass@1, the percentage of tasks whose backend passes the functional tests regardless of security, using the default prompt with no security reminder; higher is better.
Metric: Correct solutions, pass@1 (%). Source: baxbench.com. Saturation forecast: Around February 2027. 38 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.5 (Thinking) | 86.22 | #87 (Claude Opus 4.5) |
| 2 | Claude Sonnet 4 (Thinking) | 75 | #207 (Claude Sonnet 4) |
| 3 | GPT-5 | 70.66 | #105 |
| 4 | O3 | 67.6 | #132 |
| 5 | Claude 3.7 Sonnet (Thinking) | 63.27 | #259 (Claude 3.7 Sonnet) |
| 6 | O1 | 63.01 | #190 |
| 7 | O3 Mini | 63.01 | #282 |
| 8 | DeepSeek R1 | 58.42 | #262 |
| 9 | Grok 4 | 57.65 | #189 |
| 10 | GPT-4.1 | 56.35 | #258 |
| 11 | Claude 3.5 Sonnet | 55.54 | #362 |
| 12 | Qwen 3 Coder 480B A35B Instruct | 53.42 | #314 |
| 13 | Gemini 2.5 Pro | 49.77 | #159 |
| 14 | GPT-4.1 Mini | 48.49 | #363 |
| 15 | DeepSeek V3 (0324) | 48.14 | #349 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=baxbench-no-security-reminder-correct · How It Works · Data refreshed daily, snapshot 2026-11-05.